Owen Hobbs

Former Head of Trading & Product Platforms · CoinAlpha

United States · Hong Kong · Singapore

The Background

Scaled CoinAlpha's market making business from zero to 100+ concurrent markets across three regions. Owned the execution infrastructure, operational tooling, and platforms the trading team ran on.

100+ markets · 99%+ uptime · 3 regions

The Gap

Tools existed for each step individually, but connecting them into a single cohesive pipeline required domain knowledge most engineers don't have — and most traders can't build.

research → backtest → optimize → deploy → monitor

The Approach

Domain expertise + AI = force multiplier. Every architecture decision and line of code built with Claude. Not an engineer by training — an operations leader who proved the model works.

domain expert + AI = force multiplier

Built With Claude

Claude isn't just a tool I used. It's a development partner. The entire Helios system was built through continuous collaboration: designing the agent architecture, implementing the research pipeline, debugging edge cases, and iterating on this website's design.

Architecture

Agent Architecture

Three-layer monitoring system — deterministic checks, LLM analysts, rule-based router. 93% cost reduction through intelligent escalation.

Research

Research Pipeline

Autonomous research agents that test trading hypotheses through budget-constrained, goal-directed experimentation.

Frontend

This Website

Next.js frontend, Supabase data architecture, Recharts visualizations, and automated webhook pipeline for live trade ingestion.

Operations

Live Monitoring

24/7 autonomous operation with crash recovery, state persistence, and cross-agent escalation for anomaly investigation.

The Agent Team

$ helios agents --status

[LAYER 1]deterministic checks: 6 active, all passing

[LAYER 2]llm analysts: 3 on standby, 0 active

[LAYER 3]router: nominal, next cycle in 28m

bug_check

Code changes, runtime errors, stack traces — deterministic, instant, free

health_check

API connectivity, memory, process heartbeats — deterministic, instant, free

signal_check

Signal consistency, threshold drift — deterministic, instant, free

trade_analyst

Win/loss patterns, exit timing analysis — LLM, on-demand when flagged

investigation_analyst

Root cause analysis for anomalies — LLM, on-demand when flagged

strategy_analyst

Strategy health, adaptation recommendations — LLM, on-demand when flagged

Deterministic-first: Checks run every cycle at zero cost. LLM analysts are invoked only when checks detect anomalies requiring reasoning.

What's Next

Autonomous Strategy Pipeline

In Progress

I provide the hypothesis, agents handle build, backtest, optimize, validate, and deploy end-to-end.

ML Regime Detection

Planned

Machine learning agents that identify market regime changes and adapt strategy parameters automatically.

Portfolio Optimizer Agent

Planned

Autonomous cross-market allocation, position sizing, and correlation management.

Multi-Market Expansion

Planned

Extending the agent platform to manage ETH, SOL, and other markets simultaneously.

Contact

@oh92 Owen Hobbs hello@owen-hobbs.com

A Note on Data Privacy

All trade data displayed on this platform is intentionally delayed, and strategy parameters are sanitized to protect proprietary information. While the system is fully operational and trading live, the specific details shared here are designed to demonstrate capability without exposing sensitive trading logic or real-time positions.